编程

Luxury Resale Research

试用

Compare luxury resale listings and sales

它能做什么

Research public, read-only marketplace data for luxury fashion, handbags, watches, sneakers, and streetwear. Use this skill when the user needs current asks, sold comparables where the source exposes them, listing-level condition, or seller/dealer context across luxury resale sources. It complement…

技能文档

Luxury resale research

Research public, read-only marketplace data for luxury fashion, handbags, watches, sneakers, and streetwear. Use this skill when the user needs current asks, sold comparables where the source exposes them, listing-level condition, or seller/dealer context across luxury resale sources. It complements the broad resale-secondhand-research skill; use this one when specialist luxury sources or a disciplined evidence comparison matters.

Do not place orders, make offers, authenticate an item, estimate an appraisal, claim a purchase total, or imply that a platform's condition label proves authenticity. Do not infer fees, duties, shipping, or currency conversion when the selected public response does not state them.

Evidence workflow

  1. Define the comparability key first: brand, model/reference or style ID, material/color, size, condition, and region/currency where present. Search active and sold datasets separately; an asking price is not sale evidence.
  2. Discover opaque filters before searching. Use each platform's directory, categories, conditions, brands, models, or facets endpoint rather than guessing ids, slugs, or values.
  3. Pull a compact active set, then fetch detail only for credible matches. Keep source URL/path and the retrieval time for every record. Follow only the source's cursor/page limit; narrow the query instead of deep paging.
  4. Report separate fields for: active asks; sold evidence (sale price and sale date only where the endpoint supplies both); condition and authentication signals (source-stated, never independently verified); fees/shipping/duties (stated amount or not provided); currency; and source timestamp (listing/sale date if supplied, plus retrieval time). Do not pool currencies or condition grades without labeling the normalization assumption.
  5. State evidence gaps plainly. Current inventory, recommendation rails, seller reputation, and marketplace-wide counts are useful context, not a transaction record or authenticity determination.

Source map

Use the exact routes and parameter names below. All are GET public-data reads. Braces identify a path parameter; all others are query parameters.

Core luxury sources

  • The RealReal: discovery: therealreal_autocomplete (/therealreal/autocomplete, term), therealreal_categories, therealreal_conditions, therealreal_collections, and therealreal_designers (category). Search: therealreal_search (/therealreal/search, required query; optional category_id, designer_id, condition_id, price_from, price_to, on_sale, available, after). Browse: therealreal_category (path, on_sale, available, after), therealreal_designer (slug, on_sale, available, after), and therealreal_collection (slug, after). Detail: therealreal_listing (url) and therealreal_similar (product_id). Reuse next_after exactly as returned.
  • Fashionphile: discovery/catalog: fashionphile_collections and fashionphile_products (page, limit; default 50, maximum 250), plus fashionphile_sitemaps and fashionphile_sitemap_urls (type, limit). Search: fashionphile_search (/fashionphile/search; q, condition, availability, vendor, sort, page, limit; default 20, maximum 50). Use availability=sold specifically for its sold archive; available is the default and all is also accepted. Details: fashionphile_product (handle) and fashionphile_product_recommendations (handle, intent, limit; default 10, maximum 20; intent is related or complementary). Collection and static-page reads use fashionphile_collection_products (handle, page, limit), fashionphile_pages (page, limit), and fashionphile_page (handle); suggestions use fashionphile_search_suggest (q, types, limit; maximum 20).
  • Rebag: use the same public storefront pattern with only Rebag routes: rebag_search (/rebag/search; q, sort, page, limit), rebag_product (handle), rebag_product_recommendations (handle, intent, limit), rebag_products and rebag_collections (page, limit; default 50, maximum 250), rebag_collection_products (handle, page, limit), rebag_search_suggest (q, types, limit; maximum 20), rebag_sitemaps, rebag_sitemap_urls (type, limit), rebag_pages (page, limit), rebag_page (handle), and rebag_store.
  • Vestiaire Collective: discovery: vestiaire_brands, vestiaire_categories, and vestiaire_conditions. Search: vestiaire_search (/vestiaire/search; q, brand_id, category_id, condition_id, sort, page, per_page). per_page is at most 60 and page * per_page must not exceed 1000; use the returned directory IDs. Detail/context: vestiaire_product (path), vestiaire_search_sellers (q), vestiaire_seller (id), and vestiaire_suggest (q).
  • Grailed: discover filters with grailed_categories and designers with grailed_designers (q, page, limit); curated browse is grailed_collections then grailed_collection (id). Use grailed_search for active asks and grailed_sold_listings for sold evidence. Both accept q, designer, department, category, size, color, condition, min_price, max_price, sort, page, and limit; price filters mean final sold price on the sold route. Detail/context: grailed_listing (id), grailed_similar_listings (id, limit), grailed_seller (username, page, limit), grailed_seller_reviews (username, page), and grailed_suggest (q).
  • Chrono24: discover chrono24_brands, then chrono24_models (brand), and chrono24_facets (group) before advanced filters. Search with chrono24_search (/chrono24/search; at least query or brand; optional model, sort, page, page_size, condition, used_or_new, case_material, dial_color, bracelet_material, movement_type, gender, watch_type, stock_info). The UI page sizes are 30, 60, or 120. Use chrono24_listing (path) for the reference number, specifications, price/currency, condition, seller type, and location; dealer context is chrono24_dealer (slug) and chrono24_dealer_reviews (slug, page, page_size 1–50, min_stars 0–5, sort). Suggestions: chrono24_autocomplete (query).

Adjacent resale benchmarks

Use these only when they improve the requested comparison; do not substitute a streetwear marketplace for a luxury-listing comparable without saying so.

  • Depop: depop_search (query plus brand_ids, category, subcategory, condition, colours, sizes, gender, is_kids, on_sale, price_min, price_max, sort, after); discover with depop_brands, depop_categories, depop_sizes, and depop_search_facets (query). Detail/shop: /depop/item/{slug}, /depop/item/{slug}/similar (after, limit, up to 150), depop_search_sellers (query), and /depop/shop/{username} with the compatible filters. Suggestions: depop_suggest (query).
  • Poshmark: use poshmark_brands and poshmark_categories for discovery; active browse/search is poshmark_search (query, department, max_id), /poshmark/brand/{name} (max_id), or /poshmark/category/{path} (max_id). Details are /poshmark/listing/{id}; seller context is /poshmark/closet/{username} (max_id). Reuse next_max_id exactly.
  • GOAT and StockX: use product detail for market context, not a confirmed sale comp unless the response explicitly provides a last-sale field. goat_search supports query, facets, page, and limit; discover with goat_search_facets, goat_countries, and goat_curated; detail is /goat/product/{slug} (country_code). stockx_search supports category, query, facet filters, page, and limit; discover stockx_categories and stockx_brands; detail is /stockx/product/{slug}. stockx_releases (from, page, limit) is a release calendar, not resale history.

Reporting shape

For a comparison, return a compact table or JSON list with one row per source listing and include source, listing_url_or_path, status (active, sold, or market_context), price, currency, condition_text, authentication_signal, seller_or_dealer_signal, sale_date, fees_shipping_duties, source_timestamp, and retrieved_at. Leave an unsupported field null or label it not provided; never fill it with an estimate. Keep active asks and sold evidence in separate summaries before any user-requested comparison.

相关技能

从 AdMapix API 拉取广告创意、应用、榜单和收入预估等数据,原样返回结构化 JSON。

作者 fly0pants4.3k 次安装296 星标

在本地磁盘以分类纯 Markdown 文件保存需要长期留存的事实,与智能体内置记忆并存。

作者 Iván559 次安装18 星标

以 AI 机器人身份加入视频会议,提供语音、虚拟形象与屏幕共享四种模式。

作者 johnpatternai22 次安装8 星标

把自然语言描述转为结构化 JSON,并由 mcp-diagram-generator MCP 服务生成 Draw.io、Mermaid 或 Excalidraw 图表文件。

作者 nssa.io1.0k 次安装47 星标

诊断生产力系统反复失效的根因,给出最小干预——容量测算、瓶颈定位、可靠的本地记录。

作者 Iván855 次安装69 星标

crawlora-org 的更多技能

浏览全部技能

Researches Yahoo's editorial content network — Autos, Entertainment, Health, Life, News, Shopping, Sports, and Tech — via the Crawlora API, returning clean JSON. Each vertical shares a home/category story-stream plus full-article-content pattern; Yahoo Sports adds deeper sports-data endpoints (scoreboards, standings, team/player/roster, golf, MMA, motorsports, tennis, Olympics). Use when the user wants a Yahoo section's story feed, a Yahoo article's full content, Yahoo News comments, Yahoo Shopping deals/lists, or Yahoo Sports scores/standings/schedules. Yahoo Finance and Yahoo Search are covered by their own separate skills, not this one.

作者 crawlora-org2 次安装

Fetches any company's job openings straight from its Applicant Tracking System (ATS) via the Crawlora API — Greenhouse, Lever, Workday, Ashby, SmartRecruiters, Workable, Personio, Recruitee, iCIMS, Oracle Recruiting, Rippling, Pinpoint, Eightfold, Gem, UKG, and Teamtailor — returning clean JSON. Use when the user wants every open role from a specific company's ATS-hosted career page, a single job posting's full detail, or a hiring-velocity snapshot for a company, given its ATS platform and company/board slug.

作者 crawlora-org1 次安装

Researches products, prices, sellers, and reviews across major online marketplaces and big-box/specialty retailers (Amazon, eBay, Shopify stores, Shop.app, Target, Costco, Walmart, Nike, Zara, Adidas, Best Buy, Home Depot, Sephora, SHEIN, IKEA, Chewy, and more) using the Crawlora API, returning clean JSON. Use when the user asks to find a product, compare prices or sellers, track listings, or pull marketplace/retailer reviews — instead of scraping store pages.

作者 crawlora-org1 次安装

Researches hotels, flights, attractions, short-term rentals, and live events via the Crawlora API — Booking.com, Expedia, Agoda, TripAdvisor, Trip.com, Airbnb, and Ticketmaster — returning clean JSON. Use when the user wants to search or compare hotel/stay prices and reviews, look up flight options, find attractions/things-to-do or concerts/events, or research an Airbnb host or listing.

作者 crawlora-org1 次安装

Researches Zara's catalog — category taxonomy, category listings, product detail, keyword search, search suggestions, and nearby physical stores — using the Crawlora API, returning clean JSON. Use when the user asks to find a product on Zara, browse a Zara category, search Zara's catalog by keyword, pull a Zara product's colors/sizes/images, or find a nearby Zara store — instead of scraping zara.com.

作者 crawlora-org1 次安装

Researches global news and US television coverage via the Crawlora API — GDELT web-news search/context/timeline/sentiment, plus GDELT Television 2.0 AI's index of US TV transcripts, captions, on-screen text, and visual labels — returning clean JSON. Use when the user wants coverage-volume or sentiment trends over time, cross-outlet news search, sentence-level co-occurrence search, or what US TV news said/showed about a topic — OSINT, media-monitoring, and political/social-science research.

作者 crawlora-org1 次安装